The present scope of the study is to collect the cyclonic disturbances in the Bay of Bengal basin which is one of the vibrant gulfs of the North Indian Ocean (NIO). Its present geographical poisoning, vibrant physiography and stratigraphy is engulfing a lion’s share of the cyclonic disturbances formed in the NIO. The classification, naming, and the players in the process have been reported. The cyclogenesis in the Bay of Bengal, their formation processes, intensification, movement, land slamming, and their concurrent management policies are the focusing point of the research.
The soft computing study uses cyclogenesis data of 40 years starting from 1981 to 2020 based on Indian Meteorological Department publications and Wiki data. Three time-dependent observations such as CD, CS, and SCS are considered for prediction. The machine learning approach that is widely used for classification and prediction is preferred over traditional ARIMA (Auto-Regressive Integrated Moving Average). The present study makes use of WEKA, a popular software tool for the application of machine learning algorithms. The tool enables the creation of lag variables and converts the time series forecasting problem into a supervised machine learning problem. It also provides configuration options for the number of time units to forecast.
Data consideration:
The data reporting and hazard risk recording of cyclonic storms were started in 1891 through India Meteorological Department (IMD) established in India in 1875. The subcontinent has experienced 1597 cyclonic disturbances (CDs), out of which 670, and 321 numbers of CD’s have developed slammed the EC, and WC of India. The lion’s shares of 100 CS’s (19.12%) out of a total of 523 cyclonic storms have intensified from CD’s and have made their landfall along the Odisha coast till May 2021 inclusive the VSCS “YAAS” (Mishra and Ojha 2021[34]).
Past History:
As per record, the devastating intensified storm that slammed Odisha coast were 1737 (Oct), 1831 (Oct), 1864 (Oct), 1885 (22nd Sept), 1942, 1967, 1971, 1977, 1999 (Super Cyclone), 2013 (Phailin), 2014 (Hudhud), 2018 (Titli), Fani (3rd May 2019), and YAAS (26th May 2021) Sarkhel et al., 2019[35]. The cyclones are formed in-situ or remnants of typhoons of the South China Sea moved to BoB. They intensify and move towards the East coast of India. Many cyclones formed in May at times take a curvature heading towards Myanmar coast, the least formed June-Sept storms landfall across the north coast(Kolkata and Ongle ). The Oct. to Dec. storms are devastating and cross all along the east coast of India and rarely move towards Bangladesh and Myanmar coast, Fig 2 (a).
Past statistics:
The cyclones (BoB+ AS + in-land) have been considered for the periods 1891 -1949 (post-Industrial revolution and post-Holocene epoch), 1950 -1979 (global atomic activities, i.e. pre-Anthropocene period), and 1980-2020, (Golden spike period of Anthropocene when the rate of demographic growth exceeded the rate of food production in India).
The analysis of data reveals that the number of CDs and TCs formed during the post-Holocene, Pre Anthropocene, and golden spike period of Anthropocene shows changes (Fig 3, Table 2).
Table 2: The number of CD’s, CS’s and SCS’s that landfall the different coasts of India
Year
|
Cyclonic disturbances (CDs)
|
Cyclonic storms
|
Severe Cyclonic storms
|
|
BoB
|
AS
|
In-land
|
Total
|
BoB
|
AS
|
In-land
|
Total
|
BoB
|
AS
|
In-land
|
Total
|
1891-1949 (NO)
|
630
|
74
|
67
|
771
|
285
|
53
|
9
|
347
|
94
|
29
|
3
|
126
|
1891-1949 (%)
|
81.7
|
9.60
|
8.7
|
100.0
|
82.1
|
15.3
|
2.6
|
100
|
74.6
|
23.0
|
2 4
|
100
|
1950-1979 (No)
|
329
|
72
|
37
|
438
|
126
|
34
|
4
|
164
|
75
|
21
|
1
|
97
|
1950-1979(%)
|
75.1
|
16.4
|
8.5
|
100
|
76.8
|
20.7
|
2.5
|
100
|
77.3
|
21.7
|
1.0
|
100
|
1980-2020 (No)
|
261
|
89
|
38
|
388
|
111
|
47
|
1
|
159
|
67
|
31
|
0
|
98
|
1980-2020(%)
|
67. 3
|
22.9
|
9.8
|
100
|
69.8
|
29.6
|
0.6
|
100
|
68.4
|
31.6
|
0.0
|
100
|
Acronyms: No: Number; BoB: Bay of Bengal; AS: Arabian Sea
|
The % of CD’s (D + DD’s) formed for the three consecutive periods were 81.7%, 75.1%, and 67.3% during the period 1891-1949, 1950-1979 and 1980 - 2020 respectively (Fig 3 (a), (b) and (c)).
Those CD’s were intensified to CS’s and above during the same periods were also increasing gradually to 82.1%, 76.8% and 69.8% (Fig 4 (a), (b) and (c)).
But it is found that the growth from numbers of CS to higher-order storm categories like SCS, VSCS, ESCS, and SuC have shown a decreasing trend i.e. 74.6%, 77.3%, and 68.4% respectively, but with higher intensified CS’s during the same periods. Among all the cyclonic formulations, 29% of the total disturbances have affected the Odisha coast. The vulnerability of the Odisha Coast Zone is relatively high in comparison to other adjacent states like West Bengal (14%), Andhra Pradesh (13%), and Tamil Nadu (7%) (IMD Report).
Formation of cyclones in BoB:
The tropical cyclonic storm form in stages Stage I comprises of complex mesoscale vortex formation with a horizontal extension of 100 -200km and stage II is an intensification of the stage I vortex and decrease of central pressure and surge in spiral wind evolvement till cyclogenesis occur. The TC’s in BoB form when sea surface temperature is high (>26.50C), the warm water column of 50m depth (the energy source), and highly moist air rise above the ocean surface common during moths April to Dec). Rising up moist particles starts cooling and forms droplets with a dust particle as the nucleus. Condensation the droplets release heat to the atmosphere. Release and transfer of latent heat to penetrate the surrounding warm moist airs create more wind and the system grows.
The intensification of an oceanic disturbance and formation of its eye can be correlated with during formation of clouds over the sea a column of LPA (Low-pressure area) is established and create a column between the cylindrical clouds mass due to the tangential forces acting called the eye of the intensified cyclone. The continuous supply of moisture from the surroundings makes the system dynamic and makes it capable of transgression. On landfall, the moisture supply is hindered and the cyclones divulge and dissipate. The eye of an intensified storm has a diameter ranging from 30 -60 km with almost calm wind, visible blue sky, least precipitation, and fair weather with 0-20 C warmer than the surficial temperature (Gray 1975[36], Roy et al., 2012[37], Pal et al., 2020[19], Anwar et al., 2020[16],Wahiduzzaman, et al., 2021[38]).
The vertical section of an intensified cyclone extends vertically up to 12km or more where from 0-3km the movement of the cyclone occurs, 3-7km represents the body of the vorticity zone and the outflow occur beyond 7km up to 12km or more, https://www.pmfias.com/tropical-cyclones-favorable-conditions-tropical-cyclone-formation/ . A well-developed cyclonic storm should have noticeable structural elements like boundary layer, central eye and its wall, cirrus cloud shield (CDO: Cloud dense overcast), rain bands, and upper tropospheric outflow. Failure to penetrate the cloud column with sufficient wind and moisture, the cyclone shall not grow and tend to decay and die Bandyopadhyay, et al, 2020[39].
Other propelling factors are IOD, ENSO, ITCZ positioning, upper-tropospheric westerly trough, are important as they could force huge volumes of wind shear (vertical) over CD’s which may inhibit strengthening
Propelling of circulations in BoB:
After formation; the cyclone, the track may be conventional or nonconventional and at times also undergo dissipation and decay. Conventionally, the concurrent wind force propels the system to travel with sufficient vorticity forced by the Coriolis force (anti-clockwise in BoB), the centripetal force action of the earth’s rotation, and the gravitational forces, (Roy et al., 2012[37]). The conjoint factors are low altitude positive vorticity, low shear force in layers of vertical wind shell, SST (> 26.5° Celsius), surged convective uncertainty, high RH in the low and mid-troposphere (Mohapatra et al 2015[40], IMD report 2015[41]). The genesis moves forward continuously in W to NW quadrant and never takes a reverse course towards the equator. Occasionally systems remain geostatic and accumulate more energy for changing direction or recurve in the BoB. Dominantly, the systems collapse or recurvate during pre-monsoon in the Bay of Bengal, http://www.rsmcnewdelhi.imd.gov.in/images/pdf/faq.pdf. The conventional projectile of propagation in NW-ly of the cyclones is generally followed in BoB (NIO). The factors for the unconventional track are the steering Jet current, preexistence of the previous cyclone, deep unstable RH, ITCZ poisoning, high SST within the zone of confluence, strong vertical wind shear in the region, and β – effect Mohanty U C 1994[42], Mohapatra et al., 2012[43], Paliwal et al., 2014[44], https://www.indiannavy.nic.in/ifc-ior/Cyclone_Study.
Fewer cyclones during SW monsoon:
The Bay of Bengal has become a warm pull having SST up to 30-320 C (Thresh hold 26.50C for cyclogenesis) due to global warming, tectonic/volcanic activities, and Terrestrial/Oceanic Carbon Sequestration. The vast expanse, escalated temperature, RH, huge vertical wind shear make the SE and SW bay region vulnerable for cyclogenesis in BoB. The pre and post-monsoon days are conducive for the formation of disturbances and intensification due to easterly jet stream and positioning around the ITCZ (Fig 6), ENSO, and MJO activities.
The probable potential zone of cyclone genesis is North BoB during SW- monsoon. During SW-monsoon days, a monsoon trough is developed from NW India to the cyclonic disturbances in the bay till intensification to depression stage only. Such Oceanic disturbances have a short stay within the Bay. They dissipate without further intensification and cross the Odisha or WB coast and decays with a spell of heavy rain. This surged wind shear impedes the growth of cyclonic disturbances to cyclonic storms Table 3.
Table 3: The trend in the formation of CS’s in BoB during monsoon seasons during post-Holocene, Pre Anthropocene, and present golden spike period.
Year
|
JUNE
|
July
|
August
|
September
|
|
No
|
% of total CS
|
No
|
% of total CS
|
No
|
% of total CS
|
No
|
% of total CS
|
1891 - 1949
|
30
|
27.03%
|
36
|
32.43%
|
22
|
19.82%
|
23
|
20.72%
|
1950 - 1979
|
6
|
20.69%
|
5
|
17.24%
|
5
|
17.24%
|
13
|
44.83%
|
1980-2020
|
2
|
20%
|
2
|
20%
|
1
|
10%
|
5
|
50%
|
On analysis it is inferred that the trend in formation CS’s of higher frequencies observed higher during the post-Holocene period, July has the highest in number which has drastically reduced during the pre-Anthropocene period and only two during present golden spike period. Month-wise concern reveals September month is conducive for monsoon cyclones in the Bay of Bengal.
Prediction by soft computing:
A large series of data from 1891 to date is available. But data of the golden spike period (1980- 2020) has preferred to be used as there is an abrupt change in cyclogenesis after the golden spike period of the Anthropocene epoch. The conducive parameters for cyclogenesis that has changed thermal status over the Bay of Bengal by the active Barren Island volcanic activity, Sunda Island geographical turmoil, global warming, and carbon sequestration after 1980 onwards.
Two popular machine learning algorithms, Linear Regression (LR) and Sequential Minimal Optimization Regression (SMOreg), are selected as candidates for the problem at hand. SMOreg is based on support vector machine and supports both linear and nonlinear regression models.
Both the regression algorithms are set to their default configurations in the tool environment. Linear Regression uses Ridge regression to avoid over-fitting and SMOreg uses Normalized Poly Kernel as a nonlinear kernel function. The performances of both the algorithms are evaluated using the holdout method taking 70% as the training data and 30% as the test data. The RMSE (Root Mean Squared Error) is taken as an evaluation parameter to assess the accuracy of prediction. The RMSE values of both the algorithms for 10 steps ahead (2020-2030) are observed on training and test data. On average, SMOreg generates slightly less error than that Linear Regression (Table-4). Hence, SMOreg is selected for forecasting cyclogenesis in the present study.
Table-4: RMSE of Algorithms
Algorithm
|
Train
|
Test
|
LR
|
1.88048
|
2.74179
|
SMOreg
|
1.78064
|
2.56852
|